complementary dna sequencing: expressed sequence tags and ... · surfaced. proponents of cdna...

Complementary DNA Sequencing: Expressed Sequence Tags and Human Genome ProjectAuthor(s): Mark D. Adams, Jenny M. Kelley, Jeannine D. Gocayne, Mark Dubnick, Mihael H.Polymeropoulos, Hong Xiao, Carl R. Merril, Andrew Wu, Bjorn Olde, Ruben F. Moreno,Anthony R. Kerlavage, W. Richard McCombie, J. Craig VenterSource: Science, New Series, Vol. 252, No. 5013 (Jun. 21, 1991), pp. 1651-1656Published by: American Association for the Advancement of ScienceStable URL: http://www.jstor.org/stable/2876333 .Accessed: 22/03/2011 12:49

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=aaas. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve andextend access to Science.

http://www.jstor.org

http://www.jstor.org/action/showPublisher?publisherCode=aaas

http://www.jstor.org/stable/2876333?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

http://www.jstor.org/action/showPublisher?publisherCode=aaas

87. E. R. Fearon et at., Science 247, 49 (1990). 88. K. W. Kinzler et at., ibid. 251, 1366 (1991). 89. S. Srivastava, Z. Zou, K. Pirollo, W. Blattner, E. H. Chang, Nature 348, 747

(1990). 90. S. Kwok and J. J. Sninsky, PCR Technology: Principles and Applications for DNA

Ampltfication, H. Erlich, Ed. (Stockton, New York, 1989), pp. 235-244. 91. T. J. White, R. Madej, D. H. Persing, Adv. Clin. Chem., in press; S. D. Williams

and S. Kwok, in Laboratory Diagnosis of Viral Infections, E. H. Lennette, Ed. (Dekker, New York, 1991), pp. 147-173.

92. P. S. Walsh, D. A. Matzger, R. Higuchi, Biotechniques 10, 506 (1991). 93. E. Blake, J. Mihalovich, R. Higuchi, P. S. Walsh, H. Erlich, unpublished data. 94. J. Billet al.,J. Exp. Med. 169, 115 (1989); B. Budowle et al.,Am.J. Hum. Genet.

48, 137 (1991).

95. R. C. Allen, G. Graves, B. Budowle, Biotechniques 7, 736 (1989); K. Kasai et al., J. Forensic Sci. 35, 1196 (1990).

96. S. Scharf, unpublished data. 97. R. Helmuth et at., Am. J. Hum. Genet. 47, 515 (1990). 98. M. Stoneking, D. Hedgecock, R. Higuchi, L. Vigilant, H. A. Erlich, ibid. 48, 370

(1991). 99. We thank our colleagues who contributed to the development and application of

PCR. The space constraints of this review and the many publications on PCR prevent a comprehensive survey of advances and applications; we apologize to any of our colleagues whose studies have not been noted specifically. We are grateful to R. Saiki, S. Scharf, R. Higuchi, and R. Abramson for allowing us to cite their unpublished work; E. Rose for critical review; and K. Levenson for preparation of this manuscript.

Complementary DNA Sequencing: Expressed

Sequence Tags and Human Genome Project

MARK D. ADAMS, JENNY M. KELLEY, JEANNINE D. GOCAYNE, MARK DUBNICK, MIHAEL H. POLYMEROPOULOS, HONG XlAO, CARL R. MERRIL, ANDREW WU,

BJORN OLDE, RUBEN F. MORENO, ANTHONY R. KERLAVAGE, W. RIcHARD MCCOMBIE, J. CRAIG VENTER*

Automated partial DNA sequencing was conducted on more than 600 randomly selected human brain complementary DNA (cDNA) clones to generate expressed sequence tags (ESTs). ESTs have applications in the discovery of new human genes, mapping of the human genome, and identification of coding regions in genomic sequences. Of the sequences generated, 337 represent new genes, including 48 with significant similarity to genes from other organisms, such as a yeast RNA polymerase II subunit; Drosophila kinesin, Notch, and Enhancer of split; and a murine tyrosine kinase receptor. Forty-six ESTs were mapped to chromosomes after amplification by the polymerase chain reaction. This fast approach to cDNA characterization will facilitate the tagging of most human genes in a few years at a fraction of the cost of complete genomic sequencing, provide new genetic markers, and serve as a resource in diverse biological research fields.

T > HE HUMAN GENOME IS ESTIMATED TO CONSIST OF 50,000 to 100,000 genes, up to 30,000 of which may be expressed in the brain (1). However, GenBank lists the sequence of

only a few thousand human genes and <200 human brain messen- ger RNAs (mRNAs) (2). Once dedicated human chromosome

M. D. Adams, J. M. Kelley, J. D. Gocayne, M. Dubnick, A. Wu, B. Olde, R. F. Moreno, A. R. Kerlavage, W. R. McCombie, and J. C. Venter are in the Section of Receptor Biochemistry and Molecular Biology, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD 20892. M. H. Polymeropoulos, H. Xiao, and C. R. Merril are in the Laboratory of Biochemical Genetics, National Institute of Mental Health, Neuroscience Center at St. Elizabeth's Hospital, Washington, DC 20032.

*To whom correspondence should be addressed.

sequencing begins in 5 years, it is expected that 12 to 15 years will be required to complete the sequence of the genome (3). It is therefore likely that the majority of human genes will remain unknown for at least the next decade. The merits of sequencing cDNA, reverse transcribed from mRNA, as a part of the human genome project have been vigorously debated since the idea of determining the complete nucleotide sequence of humans first surfaced. Proponents of cDNA sequencing have argued that because the coding sequences of genes represent the vast majority of the information content of the genome, but only 3% of the DNA, cDNA sequencing should take precedence over genomic sequencing (4). Proponents of genomic sequencing have argued the difficulty of finding every mRNA expressed in all tissues, cell types, and developmental stages and have pointed out that much valuable information from intronic and intergenic regions, including control and regulatory sequences, will be missed by cDNA sequencing (5). However, many genome enthusiasts have incorrectly stated that gene coding regions, and therefore mRNA sequences, are readily predictable from genomic sequences and have concluded that there is no need for large-scale cDNA sequencing. In fact, prediction of transcribed regions of human genomic sequence is currently feasible only for relatively large exons (6).

On the basis of our high output with automated DNA sequence analysis of 96 templates per day and consideration of the above issues, we initiated a pilot project to test the use of partial cDNA sequences (ESTs) in a comprehensive survey of expressed genes.

Sequence-tagged sites (STSs) are becoming standard markers for the physical mapping of the human genome (7). These short sequences from physically mapped clones represent uniquely identified map positions. ESTs can serve the same purpose as the random genomic DNA STSs and provide the additional feature of pointing directly to an expressed gene. An EST is simply a segment of a sequence from a cDNA clone that corresponds to an mRNA. ESTs longer than 150 bp were found to be the most usefuil for similarity searches and mapping.

21 JUNE 1991 ARTICLES 1651

Libraries of Complementary DNAs Of the estimated 30,000 genes expressed in the human brain, as

many as 20,000 may encode low-abundance, brain-specific tran- scripts (1). The fact that up to one-fourth of all genetic diseases affect neurological functions is an indication of the diversity and impor- tance of genes expressed in the brain (8).

An assumption in our choice of cDNA libraries was that random- primed and partial cDNA clones would be more informative in identifying genes and constructing a useful EST database than sequencing from the ends of filll-length cDNAs (which contain 5' and 3' untranslated sequences) would be. By obtaining coding sequences, we hoped to take advantage of more sensitive peptide comparisons, in addition to nucleotide sequence comparisons. To discover the inherent limitations to be overcome in a large-scale cDNA sequencing project, we wanted to examine the diversity of representative cDNA libraries, identify desirable and undesirable characteristics of the libraries, and determine the information content and accuracy of single-run sequencing from both coding and flanking regions. Single-run sequencing involves performing a single sequence reaction, rather than relying on multiple, redun- dant reactions from each strand. We chose three commercial human brain cDNA libraries made from mRNA isolated from the hippocampus and temporal cortex of a 2-year-old female and from a fetal brain (9).

Single-run DNA sequence data were obtained from 609 randomly chosen cDNA clones (Table 1). Double-stranded cDNA clones in the pBluescript vector (Stratagene) were sequenced by a cycle sequencing protocol (10) with dye-labeled primers and 373A DNA Sequencers (Applied Biosystems). The average length of usable sequence was 397 bases with a standard deviation of 99 bases.

Subtractive hybridization has been used by researchers to reduce the population of highly represented sequences in a cDNA library (11, 12) by selectively removing sequences shared by another library. We tested subtractive hybridization as a way of enhancing the number of brain-specific clones in the hippocampus library by hybridizing the hippocampus library with a W138 human lung fibroblast cell line cDNA library and removing the common sequences (Table 1) (12, 13).

EST Characterization Initially, EST sequences were examined for similarities in the

GenBank nucleic acid database (14). ESTs without exact GenBank

Table 2. EST matches to human genes. Matches of at least 97% were considered to indicate that the EST corresponds directly to the human gene. Number, GenBank accession number of the matched sequence. Map positions are from (8), except where indicated. LDL, low-density lipoprotein; EST names in GenBank are the three-digit number given here preceded by "ESTOO."

EST Identification Number Map location

001 B-Actin (cytoplasmic) M10277 002 B-Actin (cytoplasmic) M10277 003 B-Actin (cytoplasmic) M10277 264 y-Actin (nonmuscle) M24241 17pll-qter 265 y-Actin (nonmuscle) M24241 17pll-qter 005 CNPase M19650 006 CNPase M19650 237 ADP/ATP translocase J03591 Xql3-q26 238 Fructose-1,6-bisphosphatase X07292 17cen-q12 239 a-2-Macroglobulin M11313 12p13.3-p12.3 240 a-Fodrin M18627 9q33-34 242 a-Tubulin K00558 t

243 a-Tubulin K00558 t

004 B-Tubulin X02344 t

244 Amyloid A4 Y00264 21q21.3-22.05

245 Apolipoprotein J J02908 8 *

246 Breakpoint cluster region X02596 22qll-ql2 251 c-erbA-a-2 J03239 3p24.3 253 Calelectrin J03578 254 Calmodulin J04046 261 Elongation factor-i a X03558 t

262 Filaggrin M24355 iq2i 263 Gs protein a subunit X04408 20qi3.2-qi3.3 266 Glial fibrillary acidic protein J04569 268 Gln synthetase Y00387

280 Hexokinase t l0p11.2 269 High-mobility group 1 protein X12597 278 LDL receptor-related protein X13916

284 Na+,K+-ATPase a subunit X04297 1p13-p1l 285 Neurofilament light chain X05608 8p21 288 Phosphoglycerate kinase L00160 Xql3 362 Ret proto-oncogene M16029 1Oq11.2 363 RhoB X06820 366 Osteonectin J03040 5q3l-q33 367 Synaptophysin (p38) X06389 Xpll.23-pll.22

*Indicates that the EST was mapped in this study by PCR. tThe human hexokinase nucleotide sequence has been published (29) but does not appear in GenBank or EMBL. This EST was initially identified by matches to the mouse and rat nucleotide sequences and the human peptide sequence. tMapping information on this isotype is not available.

Table 1. Composition of cDNA library determined by random clone sequencing. Each X ZAP library (Stratagene) was converted en masse to pBluescript plasmids, transfected into Escherichia coli XLI-Blue (Stratagene) cells, and plated on plates with X-gal, isopropyl-l-thio-3-D-galactoside, and ampicillin. A total of 1058 clones were picked at random from three human brain cDNA libraries: fetal brain, 2-year-old hippocampus, and 2-year-old temporal cortex (9). Clones selected from the hippocampus library after subtraction with the fibroblast library are listed in the "Subtracted" column. Templates for DNA sequencing were PCR products or plasmids prepared by the alkaline lysis method. About half of the templates prepared by PCR failed to yield an amplified fragment suitable for sequencing. This was primarily due to use of PCR conditions that minimized the need for further purifica- tion of the product but selected against amplification of long inserts (5 ,u of E. coli fresh or frozen overnight carrying the pBluescript plasmid, 7.5 ,uM

each deoxynucleotide triphosphate, and 0.1 uiM each primer for 35 cycles: 940C, 40 s; 550C, 40 s; 720C, 90 s). A further percentage of the PCR- generated templates failed to sequence, largely because of primer-dimer or other amplification artifacts. Qiagen columns (Studio City, California) improved the percentage of plasmid templates that yielded usable sequences from about 60% with a standard alkaline lysis protocol to over 90%. Overall, 117 PCR-generated templates and 497 plasmid templates gave usable sequences. Dideoxy chain-termination sequencing reactions were performed with fluorescent dye-labeled M13 universal or reverse primers (Applied Biosystems). After a cycle sequencing protocol (10), carried out in a Perkin-Elmer Thermal Cycler, sequencing reactions were run on a 373A automated DNA sequencer (Applied Biosystems). Some sequencing reactions were performed on an Applied Biosystems robotic workstation (28). For each column, numbers are indicated followed by percents in parentheses.

EST category Hippocampus Subtracted Fetal Temporal gory ~~~~~~~~~~~~~~~~~brain cortex

Database match-human Mitochondrial genes 48 (12.8) 10 (8.6) 3 (7.9) 6 (7.5) Repeated sequences 39 (10.4) 14 (12.2) 6 (15.8) 0 (0) Ribosomal RNA 10 (2.7) 7 (6.0) 0 (0) 11 (13.8) Other nuclear genes 32 (8.6) 7 (6.0) 4 (10.5) 0 (0)

Database match-other 32 (8.6) 7 (6.0) 5 (13.2) 4 (5.0) No database match 160 (42.8) 44 (37.9) 20 (52.6) 6 (7.5) Polyadenylate insert 53 (14.1) 24 (20.7) 0 (0) 27 (33.7) No insert 1 (0.3) 3 (2.6) 0 (0) 26 (32.5)

1652 SCIENCE, VOL. 252

matches were translated in all six reading frames, and each transla- tion was compared with the protein sequence database Protein Information Resource (PIR) and the ProSite protein motif database

Table 3. EST similarities in the GenBank and PIR databases. All significant similarities (P < 0.01) with GenBank or PIR entries are listed. Matches indicate percent identical bases for nucleotides and percent similarity (identical plus conservative substitutions) for peptides. Number indicates the accession number or locus name of the matched sequence. Abbreviations used are as follows: B, bovine; BM, Brugia malayi; BMDV, bovine mucosal disease virus; C, chicken; CE, Caenorhabditis elegans; D, Drosophila melano- gaster; E, E. coli; H, human; L, lamprey; M, mouse; N, Neurospora crassa; P, pig; PP, Pseudomonas putida; PRV, Pseudorabies virus; R, rat; S, squid; T, Torpedo caltfornica; TN, transposon Tn 4556; X, Xenopus laevis; Y, yeast; UT, untranslated; MARCKS, myristoylated alanine-rich C kinase substrate; HPRT, hypoxanthine-guanine phosphoribosyltransferase; GTP, guanosine triphosphate; LAMP, lysosomal-associated membrane protein; tRNA, trans- fer RNA; snRNP, small nuclear ribonucleoprotein; IGF, insulin growth factor; Mito, mitochondrial; DBP, albumin promoter D site-binding protein; and Pol, polymerase. EST names in GenBank are the three-digit number given here preceded by "ESTOO."

EST Description Length Match Number

Nucleotide similarities (GenBank)

247 80-87 kD MARCKS (B) 277 81.5 M24638

377 Mito ATPase B subunit (B) 421 85.1 X06088

248 p ADP-ribosyltransferase substrate (B) 256 80 M27278

256 Enhancer of split (D) 264 71 M20571

257 Kinesin (D) 263 70.4 M24441

259 Xotch (X) 435 75.4 M33874 270 B-Tubulin (H) 495 82.3 X00734 271 a-Actinin (H) 272 85 X15804 273 Apolipoprotein A-I 5'-UT (H) 110 69 M20656 274 HPRT 3'-UT (H) 85 75 M26434

275 Kruppel-related Zn2+ fingers (H) 88 67 M20678 276 LAMP-1 (H) 257 71.5 J04182 289 Aconitase (P) 318 89 J05224 293 ras-like (*) 71 74 X01669 295 IGF-binding protein 5'-UT (R) 115 77.3 J04486 299 ras-like (R) 138 57 X06889 300 RP L30 (R) 189 89 K02932 301 RP S10 (R) 273 90.8 X13549 365 UT conserved sequence element (H) 85 81 M24686 368 Electromotor neuron protein (T) 112 64 M30271 371 Maternal G10 mRNA (X) 234 80 X15243 372 Catalase T (Y) 65 72.3 X04625 374 RNA Pol II 6th subunit (Y) 216 64.7 M33924

Peptide similarities (PIR) 247 80-87kD MARCKS (B) 62 82.3 S08341 377 Mito ATPase B subunit (B) 97 92.8 S00763 249 GTP-binding protein smg p25A (B) 98 89.8 A35652 375 Genome polyprotein (BMDV) 27 74.1 GNWVBV 250 60K filarial antigen (BM) 109 78.0 A28209

252 Collagen 1 (CE) 57 57.9 A31219 255 Cadherin, neuronal (C) 42 64.3 A29964 256 Enhancer of split (D) 87 78.2 A30047 259 Notch (D) 102 72.5 A24768 260 Mobilization protein MbeA (E) 47 63.8 S04790 272 Ankyrin (H) 84 60.7 A35049 271 az-Actinin (H) 89 95.5 S05503 275 Finger protein XlcGF20-1 (X) 30 80.0 S06565 279 Elongation factor Tu (*) 24 79.2 S06703 281 Monophenol monooxygenase (M) 29 69.0 YRMSCS 282 Neurogenic receptor trkB (M) 56 83.4 A35104 283 Ul snRNP 70K protein (M) 59 57.6 S04336 286 Leu-tRNA ligase (N) 48 58.3 A33475 287 Processing-enhancing protein (N) 97 79.4 S03968 289 Aconitase (P) 106 98.1 A35544 290 Pro-rich protein (clone cP7) 56 64.3 E25372 291 NtrA (PP) 31 61.3 JG0338 292 IE180 protein (PRV) 22 86.4 EDBEIF 293 ras-like (*) 53 58.5 B34788 294 Alcohol sulfotransferase (R) 35 71.4 A33569 296 Transcriptional activator DBP (R) 39 74.4 A34894 297 Myosin heavy chain (R) 60 58.3 MWRTS 298 Protein-tyrosine phosphatase (R) 22 86.4 A34845 299 ras-like (R) 55 58.2 TVHURR 300 RP L30 (R) 58 98.3 S11622 301 RP S10 (R) 67 97.0 S01881 364 Fibrinogen y chain (L) 35 77.1 FGLMGS 257 Kinesin (S) 93 91.4 A35075 368 Electromotor neuron protein (T) 32 81.3 B33319

369 Hypothetical protein (TN) 37 64.9 JQ0431 370 Various actins (*) 37 75.7 S06062 371 Maternal G10 mRNA (X) 39 94.9 S05955 373 Hypothetical protein (Y) 24 75.0 C27061 374 RNA Pol II 6th subunit (Y) 73 90.4 B34588

*Matches with sequences from several organisms.

(14). Comparisons with the ProSite motif database were done by means of the program MacPattern from the EMBL Data Library (14a). GenBank and PIR searches were conducted with our modifications of the "basic local alignment search tool" programs for nucleotide (BLASTN) and peptide (BLASTX) comparisons (15). These modifications permit many query sequences to be automati- caily searched in a sequential fashion. PIR searches were run on the National Center for Biotechnology Information BLAST network service. The BLAST programs contain a rapid database-searching algorithm that searches for local areas of similarity between two sequences and then extends the alignments on the basis of defined match and mismatch criteria. The algorithm does not consider the potential of gaps to improve the alignment, thus sacrificing some sensitivity for 60- to 80-fold increase in speed over other database- searching programs such as FASTA (16).

Sequence similarities identified by the BLAST programs were considered statistically significant with a Poisson P-value <0.01. The Poisson P-value is the probability of as high a score occurring by chance, given the number of residues in the query sequence and the database. After the BLASTN search, 30 unmatched ESTs were compared against GenBank by FASTA to determine if significant matches were missed because of the use of BLASTN for the database search. No additional statistically significant matches were found. Statistical significance does not necessarily mean functional similarity; some of the matches reported here may indicate the presence of a conserved domain or motif or simply a common protein structure pattern. Statistically significant matches to GenBank and PIR are reported in Tables 2 and 3. The length and percent identity or similarity of each alignment is given in Table 3 to aid in evaluation of match quality.

On the basis of database searches, the 609 EST sequences were classified into eight groups as shown in Table 1. Four groups, with 197 of the sequences (32% of the total), consist of matches to human sequences: repetitive etements, mitochondrial genes, ribosomal RNA genes, and other nuclear genes. Forty-eight of the sequences (8%) matched nonhuman entries in GenBank or PIR, whereas 230 (38%) had no significant matches. The remaining 134 (22%) sequences contained no insert between the Eco RI cloning sites or consisted entirely of polyadenylate.

Table 4. Matches to the ProSite motif database. Pattern matches from the ProSite database (except posttranslational modification sites) are shown. Abbreviations used are as follows: AA, amino acyl; HIGH, motif consensus in single-letter amino acid code (30); ILGF, insulin-like growth factor; DHFR, dihydrofolate reductase; EGF, epidermal growth factor; Gal-P- UDP, galactose-l-phosphate-uridyl; C2H2, two Cys and two His residues. EST names in GenBank are the three-digit number given here preceded by "ESTOO."

Motif name EST

AA-tRNA ligase "HIGH" 094 ATP-binding site A 052,068,158,177,207,091,008,261 Carboxypeptidase/Zn2+ 112 COXi 249* Cytochrome c 060,128,120,139,279*,218,063,106 DHFR 235 Elongation factor 261 EGF 187,203 2Fe/2S Ferredoxin 067 Gal-P-UDP-transferase 101 Glycoprotein hormone 112 ILGF-binding protein 193 Leu zipper 071,072,095,055,070,106,025,200,221,

107,102,114,131,260*,290*,291*,294* 164,287*, 061,369*

Nuclear localization 182, 020,183,214, 062 Rubredoxin 226 Snake toxin 085- Zinc finger (C2H2) 188,275*

*See Table 3 for ESTs with similarity to GenBank or PIR sequences.


Table 5. Accuracy of single-run double-stranded automated sequencing. ESTs listed in Table 2 and those matching mitochondrial and ribosomal genes were aligned with sequences from GenBank with the GCG program BESTFIT. The first 85 nucleotides were the polylinker sequence that was not

aligned with the pBluescript SK reference sequence. Tabulation of errors began 15 bases into the BESTFIT alignment and thus is reported beginning with bases 101 to 200.

Bases from Mismatches- Gaps* Accuracy Aligned primer ambiguities* Insertions Deletions % bases

101-200 1.45 0.18 0.19 98.2 8800 201-300 1.72 0.25 0.11 97.9 8130 301-400 2.07 0.98 0.37 96.6 5404 >400 3.53 2.63 1.06 92.8 3197

*Error rates are reported as number of mismatches, insertions, or deletions per hundred aligned bases. "Mismatches" includes ambiguous base calls.

Thirty-six ESTs matched previously sequenced human nuclear genes with more than 97% identity (Table 2). Four of these ESTs were from genes encoding enzymes involved in maintaining meta- bolic energy, including ADP/ATP (adenosine diphosphate/adenosine triphosphate) translocase, aldolase C, hexokinase, and phosphoglycerate kinase. Human homologs of genes for the bovine mitochondrial ATP synthase Fop subunit and porcine aconitase were also found (Table 3). Brain-specific cDNAs included synaptophysin, glial fibrillary acidic protein (GFAP), and neurofilament light chain. At least six ESTs were from genes encoding proteins involved in signal transduction: 2',3'-cycic nucleotide 3'-phos- phodiesterase (CNPase) (two ESTs), calmodulin, c-erbA-ac-2, G stimulating protein (G.) et subunit, and Na+,K+-ATPase et subunit. Other ESTs were matches to genes for ubiquitous structural proteins-actins, tubulins, and fodrin (nonerythroid spectrin). Eight ESTs were from genes known to be associated with genetic disorders (8). More than half of the human-matched ESTs have been mapped to chromosomes, indicating the bias of GenBank entries toward well-studied genes and proteins.

ESTs without significant GenBank matches were also compared to the ProSite database of recognized protein motifs. Not counting posttranslational modification signatures, 54 sequences contained motifs from the database (Table 4). Some patterns are found in scores or even, as in the case of the leucine zipper, hundreds of proteins that do not share the functional property implied by the presence of the motif.

Similarities to sequences from other organisms were also detected in the BLAST searches of GenBank and PIR (Table 3). Several ESTs were similar to "housekeeping" genes, including the ribosomal

proteins (RP) SlO and L30 (in the rat) and the above glycolytic enzymes. EST00257 showed strong nucleotide sequence similarity to the squid (67.4%) and Drosophila (70.4%) kinesin heavy chain. Kinesin was first described as a microtubule-associated motor protein involved in organelle transport in the squid giant axon (17). Six oncogene-related sequences were also among the cDNA dones sequenced. EST00299 and EST00283 showed similarity to several ras-related genes, and EST00248 matched the 3' untranslated region of the bovine substrate of botulinum toxin ADP-ribosyltransferase. We also observed similarities with a Saccharomyces cerevisiae RNA polymerase subunit and Torpedo caltfornica electromotor neuron-associated protein. Two ESTs may represent new members of known human gene families: EST00270 matched the three P-tubulin genes with 88 to 91% identity and EST00271 matched ac-actinin with 85% identity at the nucleotide level.

Among the most interesting of the primary sequence relationships was the similarity of ESTs to the Drosophila genes Notch and Enhancer of split. Nucleotide and peptide alignments of EST00256 and EST00259 with the Drosophila genes are shown in Fig. 1. Both genes are part of a signal cascade encoded by the "neurogenic" genes that are involved in the differentiation of neuronal and epidermal cell lineages in the neuroectoderm of the developing Drosophila embryo (18). It has been proposed that the Enhancer ofsplit protein interacts with a membrane protein that is the product of the Notch gene to convert a developmental signal into an altered pattern of gene expression (18). EST00256 matched near the 5' end of the Enhancer of split coding sequence, away from the mammalian G protein P subunit and yeast cdc4-like elements (19). Part of the EST00259 match to Notch is in the cdclO/SW16 region that is similar to three

Fig. 1. Sequence alignments of ESTs with Droso- A C phila neurogenic genes. ESTs and EST transla- 5 CAA CCACCTCGGACTCCTGCGACCGCATCAAAGACGAATT 2 KKFREPLD ........ .. DQ1TJXR 1AADLPM

tions were aligned with nucleotide and peptide 319 CCAATCAAGTTCACCATCGCCGATACACTGGAACGCATCAAGGAGGAGTT 192 PKRQRSDPVSGVGLGNNGGYASDHTMVSEYEFADQRVWSQAHLDVVDVR.

sequences of two Drosophila neurogenic genes ~ T A C ~ hC C A C A T 3 TPG~2PITSSGLTNE

wit th GCG prga BETFT The eptid 55 TCAGCTACTGCAAGCTCAGTACCACAGCCTCAAGCTCGAWTTNCAAGT 37 xAPTPPQGEvDADWiDvNvRGPDGFTPLzMIACSGGGLETGNsEE . . E viritli t.1ie (iIi(Z>(Zi prograElrl lS' I. 'lhe peptide I I 1 I 11 1111111 1 11 11 I 11 11 :1 III I .1.:. 1:: II:-III 1:111111 ... 1 I I

alignment (30) of EST00259 with the Xotch 369 CAACTTCCTGCAGGCGCACTACCACAGCATTAAATTGGAGTGCGAGAAGC 241 ... AITPPAHQDGGKHDVDARGPCGLTPLAAVRGGGLDTGEDIENNE

product, the Xenopus laevis homolog of Notch, is 105 TGGCCAGTGAGAAGTCAGAGATGCAGCKTCACTATKTGATGTACTACGAG 85 EDAPAVISDFIYQGATCHNQTDRTGETALHLAA.VTYAIMPQGLLRP.AK

also shown. (A) EST00256 with Drosophila En- 419 TGTCGAACGAGAAGACGGAGATGCAGCGCCATTATGTTATGTACTACGAG 288 DSTAQVISDLLAQGAELNATMDKTGETSLHLAARFARADAAXRLIAGAD

hancer of split (M20571)- 69.202% identity; 1 ~13~S~G hancer of split (M20571~~~); 6.22 idniy 155 AKGTCCTACGGCTTGACCATCGAGATGCACAAACAGGCTGAGACCGTC:AA 133 MPTSGNVGR

gap. (B) EST00256 productwith the product of 1:111111 III I I I I 11 III 11 I 1111 11 III . :1 11 Drosophila Enhancer of split (M20571); 72.826% 469 ATGT GCGTGGATGCACAAGCAGACCGAGATCGCCAA 338 ANCQDNTGR

similarity; 58.696% identity, 0 gaps. (C) 205 AAGC CACCTTTCCCG

EST00259 product with the Drosophila Notch 519 GCGGCTGACACACTGAACCAGCTACTTCCATrCC.1'CAGGCCGAC

product (K03508); 60.294% similarity; 255 CACCAGCAGCAGGT

43.382% identity; 5 gaps. (D) EST00259 prod- i i 11 D uct with the Xenopus Xotch product (M33874); 568 CACCACCAGCvAGT 1 TEEPVVLPDL.DDQTXHQL HLDAADLEMKAMAPTPPQGEVD

82.143% similarity; 75.714% identity, 4 gaps. B 1814 11.1

Gaps have been introduced to increase identity and similarity (indicated by dots in lines). Num- 1 QQ?TrSDSCDR?EDE}QYS ASEKSEMXHYSXMY 50 ADCMDVNVRGPDGFTPCSTGSAPA.VISDFIYG

bers in parentheses indicate the GenBank acces- 18 GPIKTIADThIKEENFLQAYLHSIKSE Q = 1864 ADCMDVNVRGPDGFTPLMEASCSGGGLETGNSEEEEDASANMISDFIGQG

sion numbers. Symbols between lines: dashes 51 EXSYGLTIEbMHC TVA/ NGICAQVLPYLS SEHQQQVFGG 99 ATCHNQTDRTGETALHLAAVTYAIMPQG ... LLRPARMPTSGMGR

indicateidentity; doubledotsindicateasimilarity I1 1 : 1.11.:: :11111:.: I IiiiLiiLLriL 1LI 1. . :1111

score of 0.5 to 1.4; single dots represent a similarity score of 0.1 to 0.4. Scores are from pairwise alignments based on the matrix of Schwartz and Dayhoff (31) as modified in the GCG package.


Fig. 2. Chromosome 7 segregation of ESTs 6 mapped by PCR. Chro- mosomes and ESTs are as follows: 1 (293*, 012, E 4 077, 058, 079, 202,

I3 086), 2 (021, 037,234), gl 3 (248*, 257*, 274*, 2 062), 4 (009, 038), 5 1 (026, 030, 104, 123), 6 (301*, 007, 219, 023, 1 2 3 4 5 6 7 8 910111213141516171819202122X Y 356), 8 (245*, 223), 10 Chromosome (024, 197, 131), 11 (016, 111), 12 (014), 13 (372*, 273*, 200) 14 (221, 201, 008), 15 (165), 16 (373*), 17 (068), 19 (368*, 080), and X (276*). PCR conditions were as follows: 60 ng of genomic DNA was used as a template for PCR with 80 ng of each oligonucleotide primer, 0.6 unit of Taq polymerase, and 1 ,Ci of a-32P-labeled deoxycytidine triphosphate. The PCR was performed in a microplate thermocycler (Techne) under the following conditions: 30 cycles of 94?C, 1.4 min; 55?C, 2 min; and 72?C, 2 min; with a final extension at 72?C for 10 min. The amplified products were analyzed on a 6% polyacryl- amide sequencing gel and visualized by autoradiography. Asterisks indicate those ESTs with similarity to GenBank or PIR sequences (Tables 2 and 3). EST names in GenBank are the three-digit number given here preceded by "ESTOO."

cell-cycle control genes in yeast and is tightly conserved in the Xenopus laevis Notch homolog, Xotch. In Drosophila, Enhancer ofsplit is required for formation of epidermal tissue. Notch contains several epidermal growth factor-like repeats and appears to be involved in cell-cell communication during development (20).

Seven genes were represented by more than one EST. Compari- sons of all the ESTs against one another revealed two overlaps of unknown ESTs: EST00233 and EST00234 matched in opposite orientations, and EST00235 and EST00236 matched in the same orientation beginning at the same nucleotide. Five human genes were represented by more than one EST: 3-actin (three), y-actin (two), ac-tubulin (two), ac-2-macroglobulin (two), and CNPase (two).

Mapping of ESTs to Human Chromosomes We used the polymerase chain reaction (PCR) to screen a series of

somatic cell hybrid cell lines containing defined sets of human chromosomes for the presence of a given EST (21). In this process, only the hybrids that contain the human gene corresponding to the EST will yield an amplified fragment. An EST is assigned to a chromosome by analysis of the segregation pattern of PCR products from hybrid DNA templates. The single human chromosome present in all hybrids that give rise to an amplified fragment is the location of the EST.

PCR mapping has been applied to 46 clones, as summarized in Fig. 2. The EST of the human gene for apolipoprotein J (also called SP-40,40 complement-associated protein and sulfated glycoprotein 2) was localized to chromosome 8. Eleven other ESTs with Gen- Bank or PIR similarities were mapped to chromosomes. Although PCR mapping of somatic cell hybrids is relatively rapid-up to three clones can be assigned per day with a single thermal cycler-it is relatively expensive, costing about ten times as much as EST sequencing. With the same oligonucleotide primers, sublocalization can be achieved with panels of fragments from specific chromosomes or pools of large genomic clones in an analogous manner. Other mapping strategies that have been proposed are multiplex in situ hybridization, prescreening with labeled flow-sorted chromosomes, and preselection by hybridization to construct chromosome specific cDNA libraries. However, these methods are limited by the purity

of the chromosome-specific material or the specific activity necessary for detection.

Automated DNA Sequencing Accuracy and GenBank Submission

ESTs that match human sequences in GenBank are excellent tools for the analysis of the accuracy of double-strand automated DNA sequencing. Ninety EST-GenBank matches were examined for the number of nucleotide mismatches and gaps required to achieve optimal alignment by the Genetics Computer Group (GCG) program BESTFIT (22). The number of mismatches, insertions, and deletions was counted for each hundred bases of the sequence (Table 5). As expected, the sequence quality was best closest to the primer and decreased' rapidly after about 400 bases. The number of deletions and insertions relative to the GenBank reference sequence increased five- to tenfold beyond 400 bases, whereas the number of mismatches doubled. The average accuracy rate for individual double-stranded sequencing runs was 97.7% for up to 400 bases.

The minimum criteria for submission of ESTs to GenBank were that sequences be at least 150 bases in length and contain <3% ambiguous base calls. The overall accuracy of sequences submitted from each template group was at least 97%, based on matches to known human genes. Three hundred forty-eight ESTs met these criteria and were submitted to GenBank with accession numbers M61953 through M62300, inclusive. All ESTs except those matching mitochondrial or ribosomal RNA (rRNA) genes and simple repetitive elements were submitted to GenBank.

Conclusions and Prospects Single-run DNA sequencing has proven to be an efficient method

of obtaining preliminary data on cDNA clones. Our results demon- strate that sufficient information is contained in 150 to 400 bases of a nucleotide sequence from one sequencing run for preliminary identification of the cDNA and localization to a chromosome. In addition to the 35 ESTs homologous to known human genes, 48 ESTs matched sequences in GenBank or PIR with moderate to striking similarity, including high-quality matches with genes from such evolutionarily distant organisms as yeast (EST00374) and Neurospora (EST00287) (Table 3).

Two hundred thirty ESTs did not match any current database entries and therefore represent new, previously uncharacterized genes. A multitude of approaches for classifying these genes exists, including complete sequencing and expression, chromosome mapping, tissue distribution, and immunological characterization. Cur- rently unidentified cDNAs will also be classified by similarity to genes from other organisms as those sequences become available. Three ESTs reported here (EST00257, EST00259, and EST0374) were identified by similarity to sequences that have appeared since the last filll release of GenBank.

The random selection approach used here revealed an unaccept- ably large number of highly represented clones in these cDNA libraries. Over 30% of the clones from the hippocampus cDNA library consisted of rRNA, mitochondrial cDNAs, or inserts con- sisting entirely of polyA. Sixty-eight ESTs matched 12 different mitochondrial genes, including 18 matches to cytochrome oxidase I. Although elimination of these uninformative clones is a priority for developing ideal cDNA libraries, techniques to reduce repeated sequencing of clones will become increasingly important as large numbers of cDNAs are sequenced. The use of library preprocessing techniques such as subtraction, which preferentially reduces the


population of certain sequences in the library (11, 12), and normal- ization, which results in all sequences being represented in approx- imately equal proportions in the library (23), should reduce repeated sequencing of high and intermediate abundance clones and maxi- mize the chances of finding rare messages from specific cell popula- tions. In our initial experiments with subtractive hybridization of the hippocampus library with a human fibroblast cDNA library, CNPase and GFAP clones were enriched greater than tenfold and twofold, respectively. Another characteristic of the ideal cDNA library would be directional cloning so that either a coding sequence or a 3' noncoding sequence could be selectively obtained.

The EST data, in conjunction with physical mapping, will provide a high resolution map of the location of genes along chromosomes, a map that would be more costly to construct by genomic sequencing and analysis. By performing a single DNA sequencing reaction on each cDNA clone, a key piece of information was obtained for the relatively low cost of about $0.12 to $0.15 per base. The EST approach will provide a new resource for the analysis of chromosome sequence and for human gene discovery.

The screening of cDNA clones to identify the protein complement of a tissue has been explored by others to a limited extent. In 1983, Putney and co-workers sequenced over 150 clones from a rabbit muscle cDNA library and identified clones for 13 of the 19 known muscle proteins, including one new isotype, but no unknown coding sequences (24). Over 400 adult head-specific cDNA clones from Drosophila have been identified by differential screening of cDNA libraries from different developmental stages (25). Im- provements in DNA sequencing technologies have now made feasible essentially complete screening of the expressed gene complement of an organism.

In our own laboratory, the EST approach should result in the partial sequencing of most human brain cDNAs in a few years. Similar approaches begun elsewhere (26) could result in a database of most human expressed genes in less than 5 years. The presence of these minimally characterized sequences in GenBank will assist research efforts in several areas of biology. The EST database will provide identification and confirmation of coding regions in naive genomic sequences. Sublocalization of cDNAs that have been mapped to chromosomes will help define the genetic content of specific chromosomal regions and permit correlation with patterns of inheritance in genetic disease. In a related experimnent, chromosome sublocalization was the key to establishing that the -y-ami- nobutyric acid-benzodiazepine receptor 3 subunit is deleted in individuals with Angelman-Prader-Willi syndrome (27). We antic- ipate that ESTs from human brain will further the identification of genes associated with other neurological diseases and will provide a more complete view of gene expression in the brain.

REFERENCES AND NOTES

1. J. G. Sutcliffe, Annu. Rev. Neurosci. 11, 157 (1988). 2. GenBank Release 66.0, December 1990. 3. D. Baltimore, Ed., "Report of the Ad Hoc Program Advisory Committee on

Complex Genomes," Reston, VA, February 1988 (National Institutes of Health, Bethesda, MD, 1988).

4. S. Brenner, Ciba Found. Symp. 149, 6 (1990). 5. "Report of the Committee on Mapping and Sequencing the Human Genome"

(National Academy Press, Washington, DC, 1988). 6. J. Fickett, Nucleic Acids Res. 10, 5303 (1982). 7. M. Olson, L. Hood, C. Cantor, D. Botstein, Science 245, 1434 (1989). 8. V. A. McKusick, Online Mendelian Inheritance in Man, from the Welch Medical

Library, Johns Hopkins University School of Medicine, Baltimore, MD. 9. Stratagene catalog numbers 936206, 936205, and 935205, respectively.

10. Cycle sequencing was performed in a Perkin-Elmer Thermal Cycler for 15 cycles of 95?C, 30 s; 60?C, 1 s; 70?C, 60 s; and 15 cycles of 95?C, 30 s; 70?C, 60 s with the Taq Dye Primer Cycle Sequencing Core Kit protocol (Applied Biosystems).

11. D. W. Schmid and C. Girou, J. Neurochem. 48, 307 (1987); J. Fargnoli, N. J. Holbrook, A. J. Fomace, Jr.,Anal. Biochem. 187, 364 (1990); J. R. Duguid and M. C. Dinauer, NucticAcids Res. 18, 2789 (1990); F. H. Travis and J. G. Sutcliffe, Proc. Nad. Acad. Sci. U. SA. 85, 1696 (1988); K. Kato, Eur. J. Neurosci. 2, 704 (1990).

12. C. W. Schweinfest et al., Genet. Anal. Tech. Appl. 7, 64 (1990). 13. H. L. Sive and T. St. John, Nucleic Acids Res. 16, 10937 (1988). 14. GenBank Release 65.0 and collective updates through 21 January 1990, PIR

Release 26.0, and ProSite (EMBL protein motif database) Release 5.0 were used. 14a.R. Fuchs, Comput. Appl. Biosci. 7, 105 (1991). 15. S. F. Altschul, W. Gish, W. Miller, E. W. Myers, D. Lipman, J. Mol. Biol. 215,

403 (1990). 16. W. R. Pearson and D. J. Lipman, Proc. Natl. Acad. Science. U. S-A. 85,2444 (1988). 17. R. D. Vale, T. S. Reese, M. P. Sheetz, Cell 42, 39 (1985). 18. J. A. Campos-Ortega, Trends Neurosci. 11, 400 (1988). 19. D. A. Hartley, A. Preiss, S. Artavanis-Tsakonas, Cell 55, 785 (1988); C. Klambt,

E. Knust, K. Tietze, J. A. Campos-Ortega, EMBOJ. 8, 203 (1989). 20. U. Banerjee and S. L. Zipursky, Neuron 4, 177 (1990). 21. We designed oligonucleotide primer pairs from EST sequences to minimize the

chance of amplifying through an intron. Introns are less likely to be present in 3' untranslated sequences, so ESTs were examined for the presence of stop codons in each reading frame and for consensus splice junctions [J. D. Hawkins, NucleicAcids Res. 16, 9893 (1988); M. B. Shapiro and P. Senapathy, ibid. 15, 7155 (1987)]. Also desirable for mapping are 3' untranslated sequences because they are less likely to be identical in other members of a gene family or in pseudogenes. We used the primers in a PCR to amplify from total human genomic DNA templates as described in the legend to Fig. 2. If the resulting product was of the expected size, then the PCR reaction was repeated with DNA templates from two panels of human-rodent somatic cell hybrids: Bios PCRable DNA (BIOS Corporation) and NIGMS Human-Rodent Somatic Cell Hybrid Mapping Panel Number 1 (NIGMS, Camden, NJ).

22. J. Devereux, P. Haeberli, 0. Smithies, Nucleic Acids Res. 12, 387 (1984). 23. S. R. Patanjali, S. Parimoo, S. M. Weissman, Proc. Natl. Acad. Sci. U.S.A. 88,

1943 (1991); M. S. H. Koch, Nucleic Acids Res. 18, 5705 (1990). 24. S. D. Putney, W. C. Herlihy, P. Schimmel, Nature 302, 718 (1983). 25. M. J. Palazzolo et al., Neuron 3, 527 (1989). 26. Hum. Genome News 2 (no. 6), 1 (1991). 27. J. Wagstaff et al., Am. J. Hum. Genet., in press. 28. R. Cathcart, Nature 347, 310 (1990). 29. S. Nishi, S. Seino, G. I. Bell, Biochem. Biophys. Res. Commun. 157, 937 (1988). 30. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp;

E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

31. R. 0. Schwartz and M. 0. Dayhoff, inAtlas of Protein Sequences and Structure, M. 0. Dayhoff, Ed. (National Biomedical Research Foundation, Washington, DC, 1979), pp. 353-358.

32. Dedicated to the memory of J. E. Venter, 18 August 1923 to 10 June 1982.


complementary dna sequencing: expressed sequence tags and ... · surfaced. proponents of cdna...

Documents